Chap learning materialMetric guide
CRPS: scoring the whole forecast
CRPS (log1p) scores the middle line and the band together, on a scale where every district counts the same.
For Epidemiologists, Data scientists and modellers, HMIS staff
All folders / Evaluation metrics
Evaluation metrics
The six headline metrics Chap shows on every evaluation, one guide each, and a guide to reading them together.
Metric guide
MAE: cases off, on average
MAE is how many cases the middle line was off, on average. The easiest metric to explain.
For Epidemiologists, Data scientists and modellers, HMIS staff, District health officers
Metric guide
MAPE: percent off
MAPE is the miss as a share of what really happened, and why it breaks in quiet months.
For Epidemiologists, Data scientists and modellers, HMIS staff
Metrics
Reading the metrics together
The six headline metrics side by side: what each asks, when they disagree, and how to compare a model with the baseline.
For Epidemiologists, Data scientists and modellers, HMIS staff
Metric guide
RMSE: when big misses matter more
RMSE squares each miss before averaging, so one missed outbreak weighs more than many small misses.
For Epidemiologists, Data scientists and modellers, HMIS staff
Metric guide
Winkler score: tight and right
Winkler 25-75 (log1p) rewards a narrow middle band and penalises every value that lands outside it.
For Epidemiologists, Data scientists and modellers, HMIS staff
Nothing matches. Try another word or filter.